Determining the Annual Risk of TB Infection among Health Care Workers in a Public Hospital in South Africa Using the Interferon Gamma Release Assay
By
Patrick Keller MD
A Master's Paper submitted to the faculty of the University of North Carolina at Chapel Hill
In partial fulfillment of the requirements for the degree of Master of Public Health in
the Public Health Leadership Program.
BACKGROUND AND RATIONALE
Epidemiology and historical context:
Tuberculosis (TB) has been a known cause of human suffering for at least 8,000 years 1. It presents both as a chronic wasting condition and as a highly lethal epidemic. It thrives on
poverty, overcrowding, and undernutrition. While TB has been endemic in many populations
since the advent of agriculture, it rose to global pandemic levels with the industrialization of the
19th century1·23·4.
Technical advances--such as the tuberculin skin test for diagnosis of latent infection, the
advent of radiographic diagnosis, and the development of effective pharmacotherapy for
treatment--raised the hope for disease eradication. Unfortunately, human social organization
wasn't as adanced. In most of the world poverty and other social ills exacerbate poor individual
adherence to treatment and governmental failure to support control programs. This allows
Mycobacterium Tuberculosis (MTB) to develop resistance to treatment and spread uncontrolled.
The appearance and rapid success of the Human Immunodeficiency Virus (HIV) in the last 30
years has magnified the TB epidemic accelerating global spread2•
At present, TB is the leading cause of death from a curable infectious disease in adults5. In 2005 alone, TB infected 8.8 million people, and killed 1.6 million people. One third of the world's
population is estimated to be infected with TB 6 7•
MTB has multiple forms of resistance to drug therapy. Multidrug resistant MTB (MDR-TB} is
resistant to the two most important first-line drugs, Isoniazid and Rifampin. Extensively drug
resistant MTB (XDR-TB) emerged in 2006. It is resistant to first- and second-line drugs, and was
the only MTB infection growing in the United States in 20058. These varieties threaten TB
While TB remains a major public health threat, the epidemics in the U.S. and globally have
turned in the last decade. In the U.S., rates of active TB disease and TB mortality are
declining 10• Rates of MDR-TB declined in 2006 among all cases, rising only among foreign-born cases of TB. Globally, the rates of new cases and mortality are stable or declining everywhere
except sub-Saharan Africa6• 11. Even in sub-Saharan Africa more high-burden countries are
reporting to the World Health Organization (WHO) and some national control programs are
showing a reduced rate of increase in TB.
The TB pandemic, however, is not yet controlled. Drug resistant MTB varieties continue to
develop and spread in Africa and Eastern Europe, and the disease frequently crosses
international borders8• 9" 12" 13. TB and HIV stimulate one another, increasing pressure on the
fragile health systems in the world's poorest and least-stable countries 11• If TB is not controlled
in the current hot-spots a new epidemic of incurable XDR-TB may spread unchecked9· 14
TB and HIV:
The HIV epidemic drives the TB epidemic in many parts of the world, especially in
sub-Saharan Africa, where up to 70% of patients with TB are infected with HIV 15. According to the Joint United Nations Programme on HIV/AIDS (UNAIDS), 33.2 million adults and children are
living with HIV or AIDS as of 2007. HIV infection impairs immune function, which dramatically
increases the risk of developing active TB. Tuberculosis, in turn, may stimulate and activate viral
replication of HIV.16·17•
HIV weakens the immune system by causing the autolysis of macrophages and
T-lymphocyte helper cells,· the very cells responsible for the immune response to TB. In the
absence of HIV, these cells "eat" the living mycobacterium and then isolate themselves by
stimulating granuloma formation, which imprisons the mycobacteria so they do not cause
disease. Once a person is infected with TB and has undergone this immune reaction he or she
active TB disease is 10%. Active TB is most likely to occur when the body's immune quarantine
system fails, usually due to old age or additional diseases. Because HIV attacks the very cells
responsible for quarantining TB, HIV infection is a major risk factor for development of active
TB. Among people with L TBI the risk of active TB disease in individuals who also have HIV rises
from 10% per lifetime to 10% per year18 .
Conversely when MTB bacteria are consumed by an HIV infected macrophage it stimulates
the viral production of HIV. HIV in the absence of TB can also sit dormant in a host for years.
The host may go for 1 0 or more years without any symptoms and even without replicating virus.
MTB exposure to an asymptomatic person with HIV activates both diseases. When MTB infects
a previously HIV infected cell it activates the HIV genetic code in that cell to begin replicating
viruses. Macrophages produce cytokines in response to contact with MTB and these cytokines
increase the activity of HIV virus in nearby cells 18 •
This complex immunological synergy results in progressive rapid immunosuppression and
active TB disease. This can come either from activation of L TBI or rapid evolution to active TB
from recent exposure. Persons with L TBI who are infected with HIV have been shown to
develop active TB in as little as 2 years. Persons with HIV have been shown to get active TB in
early as 60 days after exposure 18. For these reasons, to ensure optimal treatment, it is important
to know the HIV status of people infected with TB and the TB status of those infected with HIV.
Measuring TB status with the TST:
The most common method for evaluating TB status in asymptomatic people is the
Tuberculin Skin Test (TST) 2· 19. The TST measures delayed type hypersensitivity skin reaction
to tuberculin proteins. It is a test of immune response to a purified protein derivative (PPD) of
non-species specific mycobacterium. The main benefits of the TST are that it is inexpensive,
The TST has a number of significant detriments. First, because the test is done in vivo,
each time someone is tested, he or she is exposed to the tuberculin PPD. With repeated testing
people can develop an immune response to the test, which causes a false positive resulf0. Second, the TST is not specific for MTB because the proteins used in the test are common to
non-tuberculous mycobacteria and proteins in the Bacillus Calmette-Guerin (BCG) vaccine2• Thus it detects lifetime cumulative exposure to all forms of mycobacteria21
• Third, immune
response to MTB wanes over time if the initial infection is isolated effectively by the immune
system. Thus in a healthy individual with L TBI there may be a false negative result on the TST.
Also, people with any immunodeficiency, such as HIV and even those with active TB disease
may receive false negative results because of reduced immune response. To partially make up
for these variations the TST is read as positive at 5mm induration in people with HIV rather then
the normal 10mm. Similarly for people in endemic areas with non-TB mycobacteria and high
rates of BCG, the recommendation is 15mm for positive6• 22 23. The TST is both non-specific and
non-sensitive in sub-Saharan Africa where community exposure is very high, non-TB
mycobacteria and BCG vaccination are common, and many people are immunocompromised
due to HIV, malnutrition, and other illnesses.
Health care workers and TB risk:
Health care workers (HCWs) are at risk of contracting MTB due to frequent or prolonged
exposure to infectious patients24• In spite of the aforementioned problems the TST is used to
estimate occupational risk of TB. One method is to use cross sectional data to evaluate the
number of people with L TBI according to a positive TST in an occupational group vs. the
general population. According to the Institute of Medicine report edited by Field some studies in
the U.S. and other high-income and low-prevalence countries using this method reported the
known confounders not accounted for in these cross sectional evaluations26. In addition to the issue of confounding, the cross sectional approach is also problematic because the TST tends
to measure lifetime cumulative exposure rather than recent exposure2. Cohort studies use TST
tests over time to identify those who convert from negative to positive. This process takes a long
time because the TST should not be repeated frequently. Other studies evaluated TB risk by
measuring the number of cases of active disease among the populations, but applying the
results from this method is complicated because active disease can occur many years after TB
infection, and both the time delay and number of cases is very dependant on the HIV rates in
the studied populations.
In 2007 Menzies published a review of TB risk among Health Care workers that included
published works from 1960 to 2005 in low and middle income countries (LMIC), and from 1990
to 2005 in high-income countries (HIC)27• It is common to separate HICs from LMICs in any discussion of TB because the risks, treatments, and recommendations vary for each group. In
HICs Menzies Joshi and Pai found that the average prevalence of L TBI for HCWs was 24%.
Positive status for L TBI in HICs was associated with non-occupational factors related to lifetime
risk such as low socioeconomic status or foreign birth. The risk of TB attributable to heath care
work in these settings was only 1.1% [0.2 -12%].
In LMICs there was a much higher burden of disease as expected. On average 63% of
HCWs in LMICs were L TBI positive. The risk of TB attributable to heath care work in these
settings was 5.8% [0 - 11 %]. In both groups the nosocomial risk was related to both the burden
of TB in the health setting and the implementation of engineering and administrative TB control
measures 27•
Earlier in the year the same authors published the review of only LMICs. In this paper Joshi
discussed the subgroups of HCWs in greater detail and presented the little data available on the
implemented. High risk of occupational exposure among these HCWs contributes to a
prevalence of L TBI ranging from 33 to 79%, and an annual risk of infection ranges from 0.5 to
14.3%. Because the numbers are so high, the attributable-risk of active TB disease due to
nosocomial exposure ranges from 25 to 5,361 per 100,000 HCWs per yea~6•
These high rates of active TB among HCWs in low- and middle-income countries are
particularly worrisome and often considered to be HIV-associated6 . The increase of disease burden from HIV and TB directly reduces the working life expectancy of health care workers in
an already severely strained environment 11. There is great concern about nosocomial risk for
TB because HCWs have frequent contact with people suffering from other chronic diseases who
are likely susceptible to TB. HCW who are not protected have the potential to spread the
disease even more than other groups if they become infectious.
Prevention and TB Infection Control:
TB Control strategy varies from country to country depending on national resources and the
burden of disease. In HICs such as the U.S. and Western Europe there is a low burden of TB.
People at risk for TB in HICs are identifiable as groups such as HCWs, the homeless,
immigrants, and others. The goal of TB control for these countries is to prevent and eliminate
the disease28. They generally use screening tests for at-risk groups and treat anyone found
positive for L TBI with preventive therapy to avoid active disease. Since Isoniazid for preventive
therapy (IPT) is inexpensive and safe, and the number of people with L TBI is small, HICs
screen with the TST to maximize sensitivity. Engineering and other measures are taken to
reduce occupational risk.
In low-income high-burden countries TB control prioritizes reduction of active transmission
of TB, not prevention29· 30• LMICs rely on identification and treatment of active cases to stop transmission. In these settings it is unclear how effective it is to treat someone with IPT because
that treating everyone is not feasible. Tests need to be highly specific to be useful and even
more so, they need to identify those most likely to be spreading the disease. For this reason,
LMICs rely on direct investigation of sputum smears for mycobacteria. People with MTB in their
sputum are most infectious and therefore receive priority treatment.
Very little is published on the effects of infection-control measures, especially in LMICs. It is
difficult for policy makers to allocate resources away from the more obvious need to identify and
treat the most infectious cases and toward preventive interventions. In order to make the
argument for prevention one must be able to estimate the nosocomial risk of TB to the workers
and the populations they serve and to demonstrate the reduction of that risk through
implementation of some or all of the available control measures.
Prevention and Occupational TB infection Control for Health Care Workers:
In 1994 the Centers for Disease Control (CDC) established comprehensive infection control
practices to protect HCWs and prevent TB transmission to the general population in U.S. health
care settings 31
• This action was taken as a response to increasing rates of nosocomial transmission and multiple TB outbreaks in the 1990s1
• 25. The CDC guidelines were not based
on evidence of effectiveness; instead they were implemented to theoretically address known
mechanisms of transmission. The guidelines include triage methods, behavioral systems of
isolation, and regular screening and treatment for L TBI, engineering procedures, such as
negative pressure rooms and antimicrobial radiation through UV lights, as well as the use of
personal protective equipment such as respirators and masks. Implementation of these
recommendations has led to a dramatic decline in the burden of TB among HCWs in the U.S.10• In order to find any information on the effects of these or other occupational infection control
programs we searched medical literature available from Medline and Google Scholar in
November 2007 for Tuberculosis or TB with combinations of secondary phrases such as health
and PLOS in order to find more recent articles. Results were limited to papers published since
1995, in which there was an evaluation of the effectiveness of TB control measures on HCWs in
a health care setting. We found five studies evaluating the impact of infection-control strategies
on the risk of TB disease among HCWs 32'36• In three of these studies the investigators (Blumberg et al, Yanai et al and Baussano et al.) evaluated reduction in TST conversion at large
referral hospitals after the full CDC recommendations for TB control (including isolation
procedures, engineering, protocol measures, and personal protective equipment for HCWs)
were implemented. Blumburg did a cohort study of house staff in the U.S. before and after
implementing TB control measures and found that the TST conversion rate dropped to less than
20% of the baseline value (5.8 down to 1.1 conversions per 100 person years (py)). Yanai et al
demonstrated a similar benefit in Thai HCWs after implementing full TB control measures (9.3
down to 2.2 conversions per 100 py) as did Baussano in Italy (2.19% per 100py down to 0.84%
per 1 OOpy). Roth et al. evaluated TST conversion in four Brazilian hospitals; two implemented
full CDC recommended TB control measures while the other two had no control measures. The
rate of conversion was more than twice as high in the hospitals with no control measures. None
of these attempted to evaluate the effect of specific parts of the overall nosocomial prevention
guidelines.
Harries et al. evaluated a less comprehensive intervention. They measured TB indirectly by
evaluating TB registries before and after limited local guidelines were imposed on hospitals in
Malawi. The guidelines were only for faster recognition and treatment of smear positive TB
cases so they were not the same as implementing the TB control guidelines. They found that
while the goal of the intervention was rapid recognition, the time to recognition and treatment
was unchanged and there was no difference in the number of HCWs being registered with TB.
One need not address nosocomial risk as an ali-or-nothing investment. It may be possible
that could be implemented at different levels of investment. These would range from the
simplest, such as outdoor waiting rooms and open windows, to more expensive interventions,
like masks for respiratory isolation or UV lights, to the most intensive interventions, such as
positive pressure rooms and full CDC guidelines. This would allow Ministries of Health to
identify which prevention measures are effective enough to complement identification and
treatment of active cases. This might be done through a cohort study that measures baseline
and follow up L TBI rates in a cohort of at-risk people (such as HCWs) over time. To improve
upon TST, an L TBI test for this purpose should be repeatable at shorter intervals and compared
with the exposure level of each individual. It should be free of interaction with non-TB
mycobacteria so that it can be accurate in a high-burden setting. It should also be independent
of BCG vaccination. Ideally, it would be more specific for recent infection than lifetime infection.
Interferon Gama Release Assays
A new group of tests for L TBI has recently been developed. These tests measure
interferon-y production by T-cells in response to in-vitro stimulation of whole blood with MTB
specific proteins not present in BCG and non-TB mycobacteria. As a group they are known as
interferon-y release assays (IGRAs)19• 37• 38. Two commercially available FDA-approved IGRAs
have been studied in recent years and are being compared with TSTs in a variety of settings38-50
. These are the Quantiferon (QFT) and Elispot tests. The IGRAs avoid some of the problems of
the TST. They use proteins that do not cross react with response to BCG vaccination or non-TB
mycobacteria. They are ex-vivo tests and so do not stimulate immunity as the TST can.
There is no gold standard for evaluating L TBI2
• 19• In order to evaluate the IGRAs, studies
have compared to TST for agreement and made correlation with expected levels of exposure, or
used active TB disease as a surrogate for L TBI. In a meta-analysis comparing the three tests to
active TB, the pooled sensitivity of the TST was 70% compared to 76% for the QFT test and
88% for the Elispot. Pooled specificity of the three tests in populations at low risk for L TBI was
to date leave more questions as to the benefits and usefulness of !GRAs for the diagnosis of
L TBI. For instance the IGRA is less likely to be positive after treatment for TB51• 52• This could represent an accurate evaluation of infection or a change in immune response. IGRA has also
shown marked variability without treatment and this should be evaluated in longer cohort studies
of untreated people that will demonstrate risk of disease for those who test IGRA positive.
To date, only a few published studies have focused on the IGRA in HCWs 52• 42· 48• 53• The studies by Neinhaus and Harada are done in high income countries (Germany and Japan) and
examine the benefits of using an IGRA in place of TST due to its possibly superior specificity for
people with BCG history.
Pai studied IGRA in a high risk environment. Working in India, Pai et al documented an
annual risk of MTB infection of 5% among HCWs, which is more than three-fold the risk in the
general population. They also observed that some HCWs had reversion of the IGRA on serial
testing. The authors suggest that transient MTB infections are occurring and can be cleared by
the immune system, a process which goes undetected with the traditional TST. Alternatively,
this could represent a high level of variability in the IGRA thus implying reduced accuracy or
reproducibility. In a study of 10 HCWs from the same cohort who were positive for L TBI and
treated with Isoniazid, Pai documented changes in the IGRA values that might demonstrate
reversion with treatment and re-conversions with continued exposure in the health care
setting51.
Corbett studied the IGRA in nursing students in a high risk setting in Harare, Zimbabwe, in
2006 54 They found an annual conversion rate with TST of 19.3 per 100 person-years in nursing students as compared to 6.0 per 100 person-years in other polytechnic students (AR 13.3, RR
3.2). In recent conference proceedings the same authors revealed preliminary results of a
comparison of TST to IGRA for evaluating conversion 55• They reported in the same group of nursing students a conversion rate by IGRA of 27.6 per 100 person-years vs the 19.3 per 100
under-estimates the TB conversion rate and that the IGRA may be a more accurate method of
assessing infection.
Summary
While TB disease rates are leveling off or declining in most of the world, with one third of
humanity infected and 1.8 million deaths a year, it remains a major public health problem. The
main threat to global public health from TB is the continually increasing number of infections and
increasing resistance to treatment in sub-Saharan Africa. Health care workers are at high risk
for TB and their loss to the infections of TB and HIV is adding further strain to an already fragile
health system. Research done thus far on the rate HCWs become infected with TB is limited,
and the difference in risk among various occupational groups is unclear. Methods of TB control
and prevention have been effective in wealthy countries and in the few middle income countries
that have implemented them in limited settings. It remains unclear, however, if these methods
can protect health care workers in high risk areas. Also unclear is which infection control
interventions, if any, carry practical benefits that outweigh their costs in high risk areas. It is too
difficult to evaluate these questions with the common test for L TBI (TST) because the TST is a
poorly effective test, both non-sensitive and non-specific in the high-risk TB settings of
sub-Saharan Africa. A new test for L TBI is now available, the IGRA. The IGRA has some qualities
that may make it a better test for evaluating TB exposure in HCWs.
This pilot study proposes to use an IGRA to estimate the annual risk of newly established
and transient MTB infections in health care workers employed in a setting with high rates of TB
and HIV. We will focus on two different groups of HCWs. First we will follow the risks of
conversion as well as reversion and variability of the IGRA as compared to TST in nurses and
technicians working in TB clinics. This is an expansion of the study of 10 HCWs done by Joshi
et al in India. We will use the data to gain a better understanding of the use of the IGRA in high
risk settings. Second we will evaluate baseline risk of conversion in students whose risk profile
working in clinical medicine and have increased patient contact. We aim to use this
methodology and baseline information to plan future studies to assess the effectiveness of TB
infection control measures in prospective controlled trials within health care settings in
sub-Saharan Africa.
SPECIFIC AIMS
Specific aim 1: Describe the fluctuations in IGRA results over a 1 year period among HCWs
involved in direct TB care.
Hypothesis: We will observe distinct patterns of IGRA responses among these HCWs: (1)
sustained high IGRA responses among HCWs with L TBI at baseline who do not receive lsonizid
Preventive Therapy (IPT); (2) conversions among HCW with negative IGRA at baseline; and (3)
reversion of IGRA response among HCW with baseline L TBI who receive I PT. A fourth possible
pattern is fluctuations of IGRA above and below the recommended threshold for a positive test.
Each of these patterns is in contrast to null pattern of no change or persistent negative test.
Rationale: Insight into IGRA responses over time among highly exposed HCW will increase
our understanding and interpretation of IGRA in this population of continuously highly exposed
individuals.
Specific aim 2: Assess the rate of IGRA and TST conversion among medical students and
nursing staff in a high prevalence area for TB and HIV. Determine the rate of agreement
between IGRA and TST conversion.
Hypothesis: The annual IGRA conversion rate will be <:1 0% among those negative at
baseline, and higher than the TST conversion rate. Agreement between IGRA and TST by
kappa score will be approximately 80%.
Rationale: Establishing an accurate annual M. tuberculosis infection rate will allow future
Specific aim 3: Evaluate knowledge and attitudes among HCWs and medical and nursing
students towards TB, L TBI, and infection control measures.
Hypothesis: Students and clinic HCWs will have good knowledge on active TB but
insufficient knowledge on L TBI and TB infection control measures.
Rationale: Insight into the current knowledge and attitudes of HCW towards the
occupational risk of M. tuberculosis infection and TB infection control measures will help guide
implementation plans for TB infection control in high risk settings.
Specific Aim 4: Estimate the prevalence of L TBI in both HCWs and medical students.
Hypothesis: HCWs will have high levels of L TBI (75%). Medical students will have low
levels (10%).
Rationale: While prevalence of L TBI has been documented in aggregated groups of HCWs
it has not been shown for physicians or medical students. Insight into the level of risk of TB
infection for different segments of the healthcare workforce may influence decisions regarding
worker protection and disease prevention.
MATERIALS AND METHODS
Setting:
This study takes place in South Africa, which according to the WHO ranks 7th on the list of
22 high TB burden countries. South Africa reported more than 270,000 new and relapsed TB
cases in 2006. This is an incidence rate of 600/100,000 population. Successful treatment
completion rates, recorded in 2004, were 70% with interruption rates of 13% and 7.4% mortality
with treatment 6.
The burden of HIV disease affects both the incidence of TB and the mortality. South Africa is
the country with the largest number of HIV infections in the world. HIV prevalence data estimate
that 29% of 15 to 24 year old pregnant women are HIV positive in South Africa56. South Africa
TB and HIV. South Africa is beset by TB as well as severe HIV epidemics and is working to
improve cure rates of active disease to 85% 30 •
Prevention of TB transmission through infection control measures is becoming increasingly
important to the South African TB strategy. As stated by acting Minister of Health, Mr. Jeff
Radebe, on World Tuberculosis Day57, "TB is one of the major health challenges currently
facing South Africa .... Realizing the challenges, the Department of Health developed the
National TB Crisis Management plan .... The Plan seeks to improve systems that are necessary
to support the TB control programme .... We have paid particular attention on strengthening
infection control precautions at our health facilities in order to reduce possible transmission of
TB in these facilities. Infection control measures are aimed at reducing direct or indirect contact
transmission by isolating patients, creating adequate bed floor space and improving ventilation
in wards."
Johannesburg is South Africa's largest city; it consists of a diverse population and a
complex public health system under Gauteng province. There are forty-eight community health
centers for the treatment of TB as well as an inpatient TB hospital for the treatment of drug
resistant varieties. Johannesburg boasts a prominent medical school and research center in the
University of Witwatersrand. In Johannesburg there is both a formidable burden of disease and
a large number of professionals and resources able to complete this study and use the results
to develop a more comprehensive study in the future.
Investigators
This project was organized and developed at UNC-CH under a grant for students and
faculty funded by the North Carolina Institute of Occupational Health and Safety to Patrick Keller
MD in the department of Social Medicine and Annelies Van Rie MD PhD in the department of
Epidemiology. Co-investigators in South Africa were recruited by Dr Van Rie through prior
contacts. These are: Charles Feldman MD PhD, Professor and Head of the Division of
Witwatersrand; Kerrigan McCarthy MBBCh DTM&H FCPath (Micro), consultant microbiologist
responsible for TB/HIV integration activities with the Reproductive Health and HIV research
Unit, Department of Obstetrics and Gynecology, University of the Witwatersrand; And Francois
Venter MD FCP, Clinical Director of the HIV Management Cluster, RHRU, and lecturer at the
Department of Medicine, University of the Witwatersrand and President of the Southern African
HIV Clinicians Society. Program design was a collaborative process between all of the
co-investigators. Management of the project in South Africa is being done by Kerrigan McCarthy.
Study Design:
This study is a prospective 1-year cohort study of two distinct populations: (1) a cohort of
health care workers involved in inpatient or outpatient TB care and (2) a cohort of medical
students beginning their clinical training.
Study Population:
Cohort 1: Approximately 50 HCWs involved with inpatient or outpatient TB care in Gauteng province. Study sites will be selected among the 48 non-mobile clinics that provide TB care and
the hospitals that provide inpatient TB care. The study sites are selected based on consultation
with the South African Ministry of Health and Department of Health for Gauteng province in
South Africa, and as a convenience sample of sites located near the University of
Witswatersrand. The actual number of volunteers recruited will depend on financial feasibility of
the study budget as well as on negotiations for permission with the National and provincial
health departments as well as individual site managers. 50 is used as a reasonable number that
can be reached. Below is a list of potential sites:
Inpatient TB care: SANTA tuberculosis hospital in Soweto, JHB; TB referral clinic at
Baragwanath Hospital
Outpatient health centers: Esselen, Jeppe, Yeoville, Urban Health, Joubert Park, and
Roodepoort, Weltevreden Park, Helderkruin, Davidsonville, Princess Clinic, Zandspruit
Clinic, Siphumlille, Nokuphila, Bophelong, ltireling, Zola, Klipspruit West, Klispruit West,
Senaoane, Tshiawelo, Protea South, Tladi, Mofolo South, Jabavu, Eastbank, Fourth
Avenue Alexandra, St Mary's, Marshalltown, Malvern, Eldorado Park, Elia Mostoaledi,
Michael Maponia, Orlando East, Noordgesis Clinic, Meadowlands, Diepkloof, Ennerdale
Ext, Lawley Ext, Lenasia Ext, Lenasia South, Mid Ennerdale.
Inclusion criteria:
HCW employed in the selected clinics
Informed consent
Exclusion criteria
Currently receiving treatment for active TB
Plans to move outside of the Gauteng area in the next 12 months
Refusal to have blood tested for HIV
Cohort 2: All students enrolled in the 3'd year in medical school at the University of
Witwatersrand will be invited to participate. These students are selected because they will be
making the transition from predominantly classroom teaching to bedside teaching and clinical
training. Students will conduct part of their training in hospital wards with high risk of
transmission of M. tuberculosis, including the Emergency and ID wards. We expect a
participation rate of 70% or 154 students enrolled at baseline. Among these, an estimated 111
students will be eligible to participate in the follow-up study.
Inclusion criteria:
Student enrolled in the 3"' year in medical school or 1st clinical year in nursing school
Exclusion criteria for enrollment at baseline:
Refusal to have blood tested for HIV
Currently receiving treatment for active TB
Exclusion criteria for participation in longitudinal study:
L TBI at baseline assessment as documented by concordant TST and IGRA test.
Recruitment and consent procedures:
Cohort 1: Following approval by the Department of Health, selected clinics will be visited by
study investigators to inform HCWs about the study and obtain informed consent of those
interested in participating. Participation is voluntary and the DOH will not be informed which
HCWs participated and which declined participation.
Cohort 2: Medical students will be recruited through an announcement at the end of a core
class by a University of the Witwatersrand faculty member not otherwise involved in the study. A
study flyer and consent form will be distributed at the end of class to interested students.
Interested students will be asked to return the completed consent form at the time of a specified
date. At that time, any questions regarding the study will be answered. The University will not be
informed which students participated and which students declined participation.
Study procedures
TST: The TST will be a single step Mantoux technique using 5 TU PPD RT23. The Mantoux
procedure is performed by injecting 5 TU of PPD subcutaneously (usually on the ventral
forearm). The procedure can be done in a single step fashion where the test is placed once and
the skin is read for reaction 48 - 72 hours later or in a two step fashion where a test is placed a
second time on those with a negative result. The two step is helpful in populations at low risk for
TB where sensitivity is to be maximized. In this case the placement of the first test stimulates
because the two-step procedure loses significant specificity without much gain in sensitivity due
to the high rates of non-tuberculous mycobacteria and BCG vaccination 58 .
Qua~tiferon Gold In-tube lnterferon-y-release-assay: The QuantiFERON ® -TB Gold assay
detects cell mediated immunity responses in-vitro to tuberculosis infection by measuring
interferon-gamma (IFN-y) harvested in plasma from whole blood incubated with the M.
tuberculosis-specific antigens, ESAT-6 & CFP-10. Five mL of venous blood will be collected and heparinized by inverting the tube several times. Aliquots of heparinized whole blood will be
incubated with ESAT-6, CFP-10, T cell mitogen (PHA) and negative control antigens. Following
16 to 24 hours incubation, the plasma will be harvested. The amount of IFN-y in the plasma
samples will be quantified by enzyme-linked immunosorbent assay (ELISA). Results are
reported in International Units relative to recombinant human IFN-y standard preparation.
In order to minimize test-related variability, identical protocols will be used for baseline and
follow-up testing, with previous results masked. All assays will be performed in the National
Health Laboratory Services, led by Wendy Stevens, who has been involved in over 600 clinical
research projects and provides laboratory support for several NIH networks (HPTN, ACTG,
PACTG, and CIPRA).
H IV test: H IV will be tested by the use of
rapid tests according to the algorithm
recommended by DOH and WHO. An HIV
diagnosis requires two positive HIV tests. The
blood sample taken at visit 1 is enough to
perform both tests.
FlOW CHART 1. ALGORITHM FOR USE Of RAPID HIV TESTS IN TESTING AND COUNSELLING SERVICES
I
NEGAlNE best «!Suit:
Cou!Utl forneg~tiw result
Negative lest result
hport result as INCOHCtVSI\IE
Reptat rapid t•utfng In &wuks
,---..J_ _ _ _ ,
N£GAnYEtenresutt:
Knowledge and attitudes questionnaire (appendix A and B): A self administered
questionnaire will be used to assess knowledge and attitudes about TB, L TBI and infection
control. The questionnaire will evaluate knowledge attitudes and beliefs among both cohorts on
TB transmission, diagnosis and treatment of L TBI, and methods for HCWs to prevent
nosocomial infection. The questionnaire will also assess TB associated stigma, and feasibility of
interventions for TB infection control. By understanding the perception of HCWs with respect to
the most feasible methods of TB infection control, we can prepare future studies of the
effectiveness of those interventions most acceptable to the local workers.
Exposure to TB among students (Appendix C): A data collection tool will be used to
document exposure to TB during work/training.
Study visits
Cohort 1: At the first study visit informed consent will be elicited in writing. Participants will
be administered the TST, receive HIV pretest counseling, and blood will be collected for the
IGRA and HIV tests. Participants will be informed that knowledge of HIV status is necessary to
correctly interpret the TST test, and that HIV results will be masked from the investigators by
separating out identifiable information. Participants will receive a questionnaire to complete (self
administered) and will be given an appointment for the reading of the TST result 48 to 72 hours
later. Participants will be encouraged to receive HIV results and posttest counseling at the time
of their choosing with a qualified councilor. Study staff will return to the clinic 48 to 72 hours later
(second study visit) to read the TST. At the second study visit staff will collect the completed
questionnaire and give information about the posttest counseling. Participants will receive an
appointment for the third study visit, 6 months later. One week prior to the third study visit, the
participant will be contacted by phone to confirm (or reschedule) the appointment. At the third
study visit, blood will be collected for IGRA and an appointment for the fourth study visit will be
confirm (or reschedule) the appointment. At the time of the fourth study visit, blood will be collected for JGRA and HIV, a TST will be placed and an appointment 48 to 72 hours later for TST reading and posttest counseling will be given. At the time of the fifth study visit, the induration of the TST test will be read and posttest counseling will be offered.
A small incentive of R70 (approximately $1 0) will be given to participants in Cohort 1 at each time a blood sample for IGRA is collected (see Figure 1).
FIGURE 1: Expected participation and results for health workers in TB clinics. 72 health workers are Results:
eligible to participate t35-TST and I or IGRA
22 (30%) are positive
expected to
*Those treated with I PT decline
participation 50 participate; will have decline then rebound of IGRA. Those including initial TST,
not treated will remain HIV and IGRA, follow
~
I · high. Those TST up IGRA at 6 months,
and final TST, HIV, negative will have high IGRA at 12 months fluctuations in IGRA
*8-HIV positive at
#-This depends on the number of clinics eligible. baseline including those
t-Assumed 70% are positive. who are TST and IGRA
:t: -Estimated based on studies by Pai et al. positive. *-Estimated at 16% based on data from Shisana et al. SAMJ 04
study visit, 6 months later. One week prior to the third study visit, participants will be contacted
by phone to confirm the appointment. At the time of the third study visit, blood will be collected
for IGRA and an appointment for the fourth study visit will be given. One week prior to the fourth
study visit, participants will be contacted by phone to confirm the appointment. At the time of the
fourth study visit, blood will be collected for IGRA and HIV, a TST will be placed and an
appointment 48 to 72 hours later for TST reading and posttest counseling will be given. At the
time of the fifth study visit, the induration of the TST will be read and posttest counseling will be
offered (see Figure 2).
I
FIGURE 2: Expected participation and results for medical students.220 students Results:
66 (30%) eligible to '21 -IGRA
likely to participate conversion
decline
'15- TST
participation
...
conversion 154 will get
initial TST, HIV,
..
*15 - HIV positive IGRA
t43 (28%) at baseline
including those
ineligible
...
who are TST andTST+ and 111 will get follow up IGRA positive.
IGRA+ IGRA at 6 months
and final TST, HIV, Unknown- HIV
IGRA at 12 months conversion rate
t · Estimated based on the data from Corbett et al.
+-Estimated IGRA and TST conversion rates 19% and 14% by averaging the data from Corbett et al. and Pai et al.
*-Estimated at 14% HIV pas. among nursing students, Cornelly SAMJ 07, 8% among medical students
I
All study visits will .be organized in groups. At each study visit, a small lunch will be provided to participants.
Care and treatment for participants with identified L TBI or HIV infection
L TBI and Preventive Therapy:
Currently, the South African department of Health only recommends TST-based screening
among people living with HIV because of the high risk of active TB in these populations30• This is in line with WHO recommendations22• 29• There is no policy regarding screening and treatment
for L TBI among HCWs. IGRA tests are available in South Africa, but test results are only used
for research purposes. Results of the IGRA will not be given out nor will they be used for clinical
decisions.
HIV seronegative individuals with conversion of TST will be informed of this result, referred
for exclusion of active TB, and advised to seek care in case symptoms of TB develop.
HIV care and treatment:
All participants will receive pretest counseling (in group for students and individual for
health care workers) prior to being tested for HIV. All study participants will also be offered
convenient post test counseling. All H IV testing and counseling will be provided free of charge
and will be performed by an NGO called New Start, which has expertise and an excellent
reputation in the field of HIV counseling and testing. Those who are found to be HIV positive will
be referred per local guidelines for HIV care and treatment.
Sample size, variables and analysis plan
Sample size:
Cohort 1: Among the HCW employed in the selected TB clinics, an estimated 50 HCWs will
participate. These HCWs are expected to have a very high rate of L TBI (75%). An estimated
15% of HCWs are expected to be HIV positive 59. Analysis of baseline results will allow estimation of prevalence of latent M. tuberculosis infection and assessment of agreement
between TST and IGRA results (Specific Aim 4). Analysis of follow-up results will provide a
determination of the distribution of patterns of IGRA responses over time and assessment of
conversion and reversion rates among heavily exposed HCWs (Specific Aim 1 ).
Cohort 2: Out of 220 eligible students an estimated 154 students will be recruited (70%).
and assessment of agreement between TST and IGRA results (Specific Aim 2 and 4). We
estimate that 28% of the students will have a concordantly positive TST and IGRA and will be
excluded from the follow-up assessments 34. Analysis of follow up IGRA and TST results in the
remaining 111 students will allow an estimation of the annual risk of MTB infection with
adequate precision (Specific Aim 2). Assuming a 15% annual risk, the sample size will result in
a 95% Cl of 0.08 to 0.22 for the annual risk of M. tuberculosis infection among medical students.
An estimated 15% of students are expected to be H IV seropositive 59•
Some have suggested that this study should be done on a larger scale to include more
students. An increase in sample size would result in a minor increase in precision of the point
estimate of the annual risk of infection, the outcome of interest. In the table below an increase in
the sample from 111 to 225 is shown as an example. We intend to limit the sample size to the
Witswatersrand medical students.
Number of 5% Loss Infection rate ULCI LLCI Number of 5% Loss to Infection rate ULCI LLCI
Participants toFU
Per vear Participants FU Per vear
111 106 10% 16 4 225 214 10% 14 06
15% 22 8 15% 20 10
20% 28 12 20% 25 15
Outcome measures:
Results of TST will be treated as a categorical variable and as a continuous measure. The
categorical response will be defined as positive if the size of induration is ~ 5mm in HIV infected
individuals and ~ 10 mm in HIV uninfected individuals. The continuous measure will be
expressed as the size of induration in mm.
Results of IGRA will also be treated as a categorical and continuous variable. The
categorical response will be defined as positive if the response is ~ 35 IU interferon gamma, as
Co variates:
Gender: Male or Female, per self report, will be treated as a categorical variable.
Age: Given per self report, will be treated as a continuous variable.
Race: Will be treated as a categorical variable the local standard classifications of race are:
White, Black, Coloured, Indian or Other. This will be defined by self report on the survey.
Baseline HIV infection status: Will be treated as a categorical variable positive or negative
per results of the two tests.
End HIV status: Will be treated as a categorical variable positive or negative.
Level of TB exposure: According to the tool developed for this study (see appendix) will be
treated as a categorical variable.
Survey responses: each of the questions in the survey will act as a separate covariate.
Some will be collated according to the categories detailed in the survey. These variables will be
treated as categorical.
Analysis plan:
Specific aim 1: Define the fluctuations in IGRA results among HCWs involved in direct TB care.
We will present the IGRA results from HCWs in graphic form baseline result, by INH
prophylaxis status and by HIV status. We will evaluate any inconsistencies with the expected 3
patterns by comparing other variables among those individuals where IGRA pattern does not
follow one of the 3 preconceived patterns.
Specific aim 2: Assess the rate of IGRA and TST conversion among medical students in a high prevalence area for TB and HIV, and agreement between IGRA and TST conversion.
We will calculate the incidence of latent M. tuberculosis infection during the 1-year follow
Concordance between TST and IGRA conversions will be evaluated using agreement and kappa statistics. We will stratify the analysis by level of exposure during the one year observation. Univariate and multivariate analysis will be used to identify variables independently associated with annual risk of infection.
Specific aim 3: Evaluate knowledge and attitudes among HCWs and medical students towards TB, L TBI, and infection control measures.
We will evaluate the knowledge and attitudes in both groups and examine if there is a difference within the two cohorts.
Specific Aim 4: Estimate the prevalence of L TBI in both HCWs and medical students.
We will calculate the prevalence of latent M. tuberculosis infection at baseline in both cohorts and provide a 95% confidence interval around the point estimate. Concordance between infection determined by TST and IGRA will be evaluated using agreement and kappa statistics. Univariate and multivariate analysis will be used to assess which variables are associated with L TBI.
Results:
At the time of this writing the study visits for Cohort 1 (clinic HCWs) have been delayed pending logistics planning and preparation in concert with our South African partners and the Gauteng Provincial Department of Health. Cohort 2 (medical students) have completed study visits 1 and 2. According to information available prior to the start of the study 220 students were expected to be eligible in the 3'd year medical school class. Instead there were 190 eligible students. Of these, only 74 (39%) participated in the study. Mean age was 23.9 years. Forty-one (55%} of the students were female. Twenty-four (32%) students were black, 31 (41%) white, 19 (26%) Indian, 3 (4%) coloured, and 2 (3%) were Asian. Most (75%) grew up in an urban
Eight students (11%) had a TST reaction greater than 15mm, 20 (27%) greater than 1 Omm, and 3 did not return for study visit 2 to have the test read. Twelve (17%) of the 70 students tested by IGRA were positive. Agreement between the IGRA and the TST with a cut off of 1 Omm was 70% (kappa 0.1508 [95% Cl: -0.08- 0.39]). Agreement between the IGRA and the TST with a cut off of 15mm was 82% (kappa 0.2997 [95% Cl: 0.06-0.54].
The association between age, gender, race, childhood environment, and parent's
occupation were evaluated with respect to TST interpretation or IGRA interpretation. Unadjusted odds ratios for each characteristic are noted in the table (Results3). None of the characteristics were associated with having a positive. People of black race were more likely to have a positive IGRA than whites (p = 0.003), and people with a rural childhood were more likely to have a positive IGRA than those with an urban upbringing (p = 0.006). Logistic regression was done with each of the variables mentioned above. Race was the only variable with a statistically significant (p=<0.05) coefficient. The results were tested with the likelihood ratio test. All variables other than race failed to add any explanatory power.
Results 1: Medical
Demographics Students
(n = 75)
Age: 23.9 (21-32)
Gender:
Female 41 (54.7%)
Male 34 (45.3%)
Race:
Black .23 (30.7%)
White 29 (38.7%)
Coloured 2 ( 2.7%)
Indian 19 (25.3%)
Other 2 ( 2.7%)
Childhood:
Urban 54 (75.0%)
Rural 9 (12.5%)
Mixed 9 (12.5%)
Were Parents Health Workers?
Neither 56 (74.7%)
Mother only 10(13.3%)
Father only 8 (10.7%)
Both 1 ( 1.3%)
TB exposure history
None 45 (60.0%)
A friend had TB 19 (25.3%) A close relative had TB 10 (13.3%)
I have had TB 1( 1.3%)
Results 2: Medical Students Visit 1 Testinq (n = 75*)
TST (in mm) Mean 5.4 SD 7 (0-30) > 10 mm 20 (27.4%)
> 15 mm 8 (11.1%)
IGRA Adjusted Mean 0.21 SD 1.8 (-1 0.53- 6.26) IGRA positive 12 (17.4%) HIV positive None
Results 3: Unadjusted Odds Ratios for IGRA results and TST results
IGRA
TST
OR
P-value
95%CI
OR
P-value
95%CI
Age
1.20
0.137
0.944-1.52
0.98
0.901
0.66-1.43
Female Sex
0.686
0.526
QJ.80-2.63
2.83
0.198
0.485-29.60
Race
White
1
Ref
1
ref
Black
28.00
0.003
3.17-247.4 5.25
0.058
0.946-29.20
Coloured
14.00
0.097
0.62-317.4 Too few
Indian
3.29
0.345
0.28-39.14 0.78
0.842
0.066-9.217
Other
Too few Too few
Too few
Childhood environment
Urban
1
Ref
1
ref
Rural
14.0
0.006
2.15-91.11
0.907
0.932
0.097-8.46
Mixed
2.33
0.352
0.39-13.91
1.16
0.894
0.122-11.2
Parent occupation as health care worker
Neither
1
Ref
1
ref
Mother
1.19
0.838
0.216-6.59
2.55
0.308
0.421-15.45
Father
2.04
0.459
0.443-9.47 2.91
0.249
0.472-17.99
Both
Too few
Too few
TB exposure history
None
1
Ref
1
ref
Friend
2.57
0.128
0.762-8.68
1.37
0.690
0.295-6.343
Family
Too few
1.03
0.983
0.105-9.989
Self
Too few
Too few
Data from the questionnaire are not yet available for analysis. Specific demographic information from the entire medical class to compare participants with non-participants is also not yet available.
The main findings of the preliminary results of this study are: 1) Far fewer medical students chose to participate than expected (39% as compared to 70%); 2) Among those who
Discussion:
In this study of medical students in Johannesburg, South Africa, a city with one of the highest TB incidence and HIV prevalence globally, we found that only 11% to 17.4% of medical students had evidence of latent tuberculosis infection when measured using TST and IGRA, respectively, and none of the participating medical students had a positive HIV test result.
Prior studies on TB in health workers from low and middle income countries show very high average rates usually above 60% 13• 26· 27· 62. None of these studies specifically evaluated medical students but the Review by Joshi et al in 2006 presented combined numbers for medical and nursing students in multiple low and middle income countries. These rates were highly variable 7% - 40%. The only African country represented was Uganda and that was a 40% L TBI prevalence.
Two articles have been published since 2003 that measure HIV rates in South African health care workers 59' 61. Neither of these studies included or identified medical students. Shisana et al. evaluated health care workers across South Africa in 2004. They used cluster methods to sample a representative 5% of the total health facilities in South Africa and surveyed and tested 595 out of 721 eligible health workers identified by their sample. They presented a total HIV rate of 15.7% for all HCWs. They specified their findings based on professionals (Physicians and Nurses, n = 349) and non-professionals (n= 246). They did not specify how many of the professionals were physicians, and they did not test students. The HIV rate for professionals was 11 .7%. Shisana et al stratified HIV results by age for the full sample. HIV prevalence was higher at younger ages. Among 18-35 year aids HIV prevalence was 20% (n = 203), among 36-45 year aids it was 16.6% (n=221 ). Based on this study we expected a 15% HIV prevalence among medical and nursing students combined. This estimate is based on the 11.7%
prevalence for professionals and the 20% prevalence for younger health workers.
along with medical students. This was changed because of practical limitations at the study site. It may be that there is a difference in the HIV risk of physicians as opposed to nurses or of students as opposed to professionals.
A study by Connelly et al was published since we developed our protocol. They tested 1493 out of 1813 HCWs at two hospitals in the Gauteng province of South Africa. They also stratified H IV prevalence by age but chose smaller intervals, 18 - 24 years, 25 - 34 years, etc. The overall HIV positive rate was 11.5%. For those 18-24 years old the prevalence was only 6.7% (n = 1 05). For 25- 35 year olds HIV prevalence was 15.9% (n = 326) The prevalence in these two age groups combined is 13.6%, this is lower than the 20% calculated by Shisana and may be more relevant for our population because it is in Gauteng Province. It also shows that the younger group, aged 18-24, is much less likely to be positive than the group of 18-35 year olds.
The Connelly study also presented HIV rates for doctors, nurses, and student nurses separately. Physicians had a surprisingly low HIV prevalence at 2% (1 out of 49). Student Nurses were the professional group with the highest HIV prevalence at 13. 8% (n=65). It is interesting to note that while they had a response rate of 82.3% among all HCWs, among physicians it was less than 25%. This is consistent with our result of 39% participation and 0% HIV positivity. It may be that physicians and physician students are at much lower risk than other health workers in South Africa. It may also be that Connelly's study and ours are suffering from the same biased selection of only the lowest risk individuals.
Race was correlated with positive IGRA results, with Black race being associated with having a positive result. Childhood background was also associated with the IGRA with people from rural environments more likely to have a positive IGRA. These results are all limited by the small numbers in this study. Other studies have shown similar results for race but usually children from urban environments are more likely to test positive for TB 2•4• 60•
involved in this study have started a student project to elicit reasons for participation or lack of participation from their peers. They hypothesize that the requirement of HIV testing is frightening and unacceptable for many because of the severe consequences that a positive result can entail in South Africa. If they are right and at-risk physicians and physician students
systematically avoid any study that could identify HIV then it may be difficult to evaluate HIV or TB in this population. Others speculate that physicians in South Africa do not carry the risk of TB and H IV infection that other health workers do because of social separation and a distant approach to the practice of medicine. Regardless of the answer, for the purpose of this study follow up studies of HIV and TB risk for the goal of testing TB control interventions may have better results with nursing students and nurses as they are more likely to participate and more likely to suffer the highest burden of disease.
As we move forward with the analysis and further data are collected it will be important to evaluate any differences between participants and non-participants to judge how well our
sample represents the medical students. Based on our data and information from Connelly, data from nursing students and other health workers will need to be analyzed separately from that of the medical students.
REFERENCES
1. Reichman LB, Hershfield ES. Tuberculosis :A Comprehensive International Approach. Vol 144. 2nd , rev. and expanded. New York: Dekker; 2000.
2. Raviglione MC, Reichman LB, Hershfield ES. Reichman and Hershfield's Tuberculosis :A Comprehensive, International Approach. Vol219. 3rd ed. New York: lnforma Healthcare; 2006.
3. Frieden TR, Sterling TR, Munsiff SS, Watt CJ, Dye C. Tuberculosis. The Lancet. 2003;362:887-899.
4. Packard RM, NetLibrary I. White Plague, Black Labor. Berkeley: University of California Press; 1989.
5. Dye C. Global epidemiology of tuberculosis. Lancet. 2006;367:938-940.
http://www.who.int.libproxy.lib.unc.edu/tb/publications/global_report/2007/pdf/full.pdf. Accessed 3/31/2007.
7. Corbett EL, Watt CJ, Walker N, et al. The growing burden of tuberculosis: Global trends and interactions with the HIV epidemic. Arch Intern Med. 2003;163:1009-1021.
8. Centers for Disease Control and Prevention (CDC). Worldwide emergence of mycobacterium tuberculosis with extensive resistance to second-line drugs. MMWR. 2005;55:10.
9. Singh JA, Upshur R, Padayatchi N. XDR-TB in south africa: No time for denial or complacency. PLoS Med. 2007;4:e50.
10. CDC. Reported tuberculosis in the united states, 2006. Atlanta, GA: U.S.: Department of Health and Human Services, CDC; 2007.
11. World Health Organization. The world health report 2006 -working together for health. Geneva, Switzerland: WHO Press; 2006. Available from:
http://www.who. intlwhr/2006/whr06 _ en.pdf. Accessed 2/29/2008.
12. Tanzania Tuberculin Survey Collaboration. Tuberculosis control in the era of the HIV epidemic: Risk of tuberculosis infection in tanzania, 1983-1998. lnt J Tuberc Lung Dis. 2001;5:103-112.
13. Naidoo S, Jinabhai CC.
TB
in health care workers in KwaZulu-natal, south africa. lnt J Tuberc Lung Dis. 2006;10:676-682.14. Moll T. TB on the back burner, losing curable status. /nt J Tuberc Lung Dis. 2007; 11 :355.
15. Gandhi NR, Moll A, Sturm AW, et al. Extensively drug-resistant tuberculosis as a cause of death in patients co-infected with tuberculosis and HIV in a rural area of south africa. Lancet. 2006;368:1575-1580.
16. Gandy M, Zumla A. The Return of the White Plague :Global Poverty and the 'New' Tuberculosis. London; New York: Verso; 2003.
17. Cole ST. Tuberculosis and the Tubercle Bacillus. Washington, DC: ASM Press; 2005. 18. Schlossberg D. Tuberculosis & Nontubercu/ous Mycobacteria/Infections. 5th ed. New York: McGraw-Hill, Medical Pub. Division; 2006.
19. Foundation for Innovative New Diagnostics, Special Programme for Research and Training in Tropical Diseases, World Health Organization. Diagnostics for Tuberculosis :Global Demand and Market Potential. Geneva: WHO on behalf of the Special Programme for Research and Training in Tropical Diseases; 2006.
21. Mahomed H, Hughes EJ, Hawkridge T, et al. Comparison of mantoux skin test with three generations of a whole blood IFN-gamma assay for tuberculosis infection. lnt J Tuberc Lung Dis. 2006;10:310-316.
22. World Health Organization. Guidelines for the prevention of tuberculosis in health care facilities in resource-limited settings. Geneva: World Health Organization; 1999.
23. Menzies R. Chapter 12 tuberculin testing. In: Reichman L, Harshfield E, eds.
Tuberculosis, A Comprehensive International Approach. Basel, Switzerland: Marcel Dekker AG; 2000:310.
24. Sepkowitz KA. Tuberculosis and the health care worker: A historical perspective. Ann lnt Med. 1994;120:71-79.
25. Field MJ, Institute of Medicine. Tuberculosis in the Workplace. Washington, D.C.: National Academy Press; 2001.
26. Joshi R, Reingold AL, Menzies D, Pai M. Tuberculosis among health-care workers in low- and middle-income countries: A systematic review. PLoS Med. 2006;3:e494.
27. Menzies D, Joshi R, Pai M. Risk of tuberculosis infection and disease associated with work in health care settings. lnt J Tuberc Lung Dis. 2007; 11 :593-605.
28. Joint statement of the american thoracic society and the centers for diseases control and prevention targeted tuberculin testing and treatment of latent tuberculosis infection. Am J Respir Crit Care Med. 2000;161 :221-247.
29. World Health Organization. Treatment of tuberculosis guidelines for national programmes. Geneva: World Health Organization; 2003. Available from:
http://whqlibdoc.who.int/hq/2003/WHO _CDS_ TB _2003.313 _ eng.pdf.
30. The South African National Tuberculosis Control Program. Practical guidelines. 2004 Accessed 3/31/2007.
31. Taylor Z, Nolan C, Blumberg H. Controlling tuberculosis in the united states,
recommendations from the american thoracic society, CDC, and the infectious diseases society of america. MMWR; 2005;54(RR12):1-81.
32. Blumberg HM, Sotir M, Erwin M, Bachman R, Shulman JA. Risk of house staff tuberculin skin test conversion in an area with a high incidence of tuberculosis. Clin Infect Dis.
1998;27:826-833.
33. Yanai H, Limpakarnjanarat K, Uthaivoravit W, Mastro TD, Mori T, Tappero JW. Risk of mycobacterium tuberculosis infection and disease among health care workers, chiang rai, thailand. lnt J Tuberc Lung Dis. 2003;7:36-45.
